Deep Learning-Based CSI Prediction Framework for Channel Aging Mitigation in TDD 5G Systems

TL;DR

This paper introduces a deep learning-based framework using LSTM networks to predict future CSI in TDD 5G systems, effectively mitigating channel aging effects caused by user mobility and delays.

Contribution

It proposes a novel LSTM-based CSI prediction method operating in the effective SINR domain, compatible with existing standards and improving link performance.

Findings

LSTM predictor improves NMSE by up to 2 dB

Achieves up to 1.2 Mbps throughput gain

Effective in moderate Doppler conditions

Abstract

Time division duplexing (TDD) has become the dominant duplexing mode in 5G and beyond due to its ability to exploit channel reciprocity for efficient downlink channel state information (CSI) acquisition. However, channel aging caused by user mobility and processing delays degrades the accuracy of CSI, leading to suboptimal link adaptation and loss of performance. To address this issue, we propose a learning-based CSI prediction framework that leverages temporal correlations in wireless channels to forecast future signal to interference plus noise ratio (SINR) values. The prediction operates in the effective SINR domain, obtained via exponential effective SINR mapping (EESM), ensuring full compatibility with existing 5G standards without requiring continuous pilot signaling. Two models are considered: a fully connected deep neural network (DNN) and an long short-term memory (LSTM)-based network. The simulation results show that the LSTM predictor achieves an improvement of up to 2 dB in normalized mean squared error (NMSE) and a gain of up to 1.2 Mbps throughput over a baseline without prediction under moderate Doppler conditions. These results confirm the potential of lightweight AI-based CSI prediction to effectively mitigate channel aging and enhance link adaptation in TDD 5G systems.